Introduction
Despite a number of high-profile international climate negotiations since 1992 (with the most recent concluding in Bonn just a few weeks ago), the climate crisis continues to deepen, and the current combined pledges toward mitigation and adaptation are woefully inadequate. With this in mind, I have created visualizations related to data on anthropogenic climate change, as this topic is highly relevant, pressing, and ripe for exploration. Rather than focusing on climate change itself (e.g. temperature and weather patterns), these visualizations explore the socio-political landscape of climate change, highlighting disparities in responsibility and vulnerability. In addition to demonstrating these inequities in the present, I aim to show (one aspect of) their historical roots as well as indications of their likely continuation in the foreseeable future.
The idea for these visualizations came out of a conversation with a professor of environmental ethics about available teaching tools that address these aspects of climate change. In particular, we discussed the kinds of elements and narrative that would be especially useful for a professor and students in this kind of class. As such, the primary target audience for these visualizations are individuals who are interested in and have at least a minimal background knowledge of climate change, and are looking to learn more about its relationship with social, political, and economic realities. However, while that is my intended audience, my hope is that these visualizations would also be accessible to the general public – that users do not need to have much background knowledge and could get something out of it even on a more superficial level.
Although several similar visualizations already exist, they can often be a bit overwhelming, even for those familiar with the concepts, vocabulary, and units of measurement, just because there are so many complex and dynamic aspects to the issue (see the Global Carbon Atlas for an example). The Carbon Map is an example of a (cartogram) visualization that is fairly easily digestible while also being comprehensive. While this served as one of the primary models for my visualizations, I did not simply want to replicate it, and have aimed to include some additional distinguishing elements.
Materials & Methods
Data Sources/Formatting & Visualization Design
The data used for these visualizations was selected for specific reasons and pulled from a variety of sources. These are broken down by each visualization section (Present, Past, and Future) below, along with rationales for the design choices made.
In addition to providing information that directly related to climate change responsibility and vulnerability, I wanted to also include some somewhat more indirectly (but still highly relevant) contextual information, such as wealth and population. To demonstrate the current concentration of wealth on an international scale (and, ultimately, how that relates to climate change responsibility), the total gross domestic product (GDP) for each nation from 2016 was included. This data was taken from the US Department of Commerce, Bureau of Economic Analysis. Initially, I had hoped to include GDP per capita, as well as greenhouse gas (GHG) emissions per capita, but for reasons I will discuss a bit more below, I decided to just include numbers on a national level to tell an overall more globally-oriented narrative. The population in each country from 2016 was pulled from The World Bank.
To demonstrate each nation’s (more-or-less) current responsibility for contributing to climate change, I included data from the World Resources Institute (WRI) Climate Analysis Indicators Tool (CAIT), which “draws on key climate-relevant data from respected research centers, government agencies, and international bodies.” According to the CAIT website, “to the extent possible, CAIT includes emissions from all greenhouse gases and major emission sources for each country” which may involve “as many as five GHG data sources.” Thus, each number for each country included in the map visualization here is the sum of available greenhouse gas emissions data from individual sectors (including emissions and removals from land-use change and forestry) and gases (i.e. CO2, CH4, N2O, and F-gases). These total emissions are measured in million metric tons of carbon dioxide equivalent (MtCO2) – a metric measure used to compare the emissions from different greenhouse gases.
While there are many different kinds of datasets related to emissions, I chose this one because it was the most comprehensive I could find; however, because of this, it is slightly old (2014), as it is the most recent data of this kind available, but I think still recent enough to be included in the “Present” section. It is worth noting that while the inclusion of land-use change and forestry data is important (it factors in carbon emissions as well as removals from these sectors) and is a more holistic approach and in theory provides a more accurate number, there are still “high data uncertainties” associated with it. It is also worth noting that these national total GHG calculations are very sensitive to data availability. Still, these risks seemed worth taking given the large scale of the visualization and overall narrative, which would still likely accurately reflect general trends. As mentioned above, I had initially wanted to include data on GHG emissions per capita (by dividing the totals by the population), but because of the way the scaling worked in Tableau, it did not produce a meaningful visualization; compared with the range of numbers from the other dataset filters, the GHG per capita numbers were relatively too close together (in any unit), and the circles representing these numbers therefore did not show sufficient variation in size to be visually useful.
To demonstrate which populations would be most vulnerable to climate change, I followed The Carbon Map’s lead and included data on people who have been and would likely be exposed to extreme weather and temperatures, people who live less than five meters above sea level and are therefore vulnerable to sea level rise, and people living on less than $1.90 a day and therefore less able to easily adapt to and cope with the changing environment. To calculate a current rough estimate of the total number of people in each of these cases per country, I used percentages of the population affected provided by The World Bank and multiplied them by the 2016 population (again, following The Carbon Map’s lead). In the case of people exposed to extreme weather, the original dataset provided average percentages of the population exposed between 1990-2009. Likewise, for people living below a five-meter elevation, the original dataset provided percentages of each country’s population from 2010. Finally, for people living in poverty, percentages of each country’s population was taken from the most recent year for which data was available. For each of these, I chose to include the number of vulnerable people (rather than percentage of the population) to give a better idea of how many actual humans are affected. While the calculations are certainly not perfectly precise, I think they provide a sufficiently accurate estimate on a global scale.
In order to include all of these datasets on one map and have the ability to filter by each of them, I first compiled all of the data in one table. However, the format of the data as it was provided was not appropriate for Tableau to read it in a way that would allow me to create filters by type of data (each type was its own column). I therefore had to use OpenRefine to “transpose cells across columns into rows.” Once the data was appropriately formatted and cleaned (e.g. country names edited to be consistent with datasets for other visualizations), it was imported into a single sheet in Tableau to create the map visualization.
As all of this data is spatially/geographically-based, using a map to display it was a natural choice. I chose to create a graduated symbol map (using circles as symbols), that allowed users to filter by the above data (i.e. population, wealth, emissions, and the three sets of vulnerability data). A graduated symbol map was preferable to a choropleth map, as the former “avoids confounding geographic area with data values and allows for more dimensions to be visualized (e.g. symbol size, shape, and color)” (Heer et al., 10-11). As a user switches between filters/datasets, the circles over each country change in size to represent that country’s total number.
Taking further advantage of pre-attentive processing, I also chose to emphasize the countries with the highest emissions by assigning them colors that would be consistent throughout the other visualizations (the rest of the countries were assigned a neutral dark gray). In accordance with the best practices for color, I limited the number to seven, and tried to avoid placing strong contrasting colors next to each other (although since the data can be filtered, this is not possible to fully control). As I was not satisfied with the colorblind-safe palette provided by Tableau, I used one generated by ColorBrewer, which I added to the options in Tableau via the Tableau Preferences file (edited in TextWrangler). I felt these colors (particularly the reds and oranges) were more appropriate for the (dire) subject matter; as Sula points out, “emotions have been found to play an important…role in information processing generally” and color is “widely regarded as having emotional connotations” (29). Although I initially thought it was obvious which colors corresponded to which countries (based on where they were located on a map), I ultimately included a legend (limited to the top emitters on the dashboard, but users can scroll through the countries if desired), which was prompted by user experience research (discussed more below) and backed up by MacDonald who recommends “always includ[ing] a color key or scale with a color-coded display” (31). This legend corresponds to the other two visualizations as well, and was therefore placed in a somewhat central location on the dashboard. Additionally, I slightly reduced the opacity of the circles on the map so that users are able to see beneath them, as smaller circles can otherwise often be obscured by nearby large circles.
To give a sense of the historical trends of how climate change responsibility breaks down by country, I created a stacked area graph of time-based data that does just that. It has been sorted by total emissions, with the highest emitters at the top. The data is from Global Carbon Atlas. I chose to include data on territorial emissions as well as consumption emissions (measured in million metric tons of CO2) because they demonstrate different ways of measuring emissions and have different histories. Territorial-based accounting measures emissions within a geographical system boundary and only includes GHG emissions generated by domestic production without consideration for where the product or service is consumed. Consumption-based accounting measures emissions within a product-oriented system boundary and considers all emissions related to the product’s life cycle. However, consistent data for consumption emissions is not available before 1990. I also chose to include data on emissions per capita here (measured in metric tons of CO2), since I was unable to on the map and it is a different and important perspective on the data.
As with the map visualization, in order to include all of these datasets on graph and have the ability to filter by each of them, I first compiled all of the data in one table and reformatted it in OpenRefine (by transposing cells across columns into rows). Once the data was appropriately formatted and cleaned (e.g. country names edited to be consistent with datasets for other visualizations), it was imported into a single sheet in Tableau to create the graph. The colors used for this graph are consistent with those used for the map.
To give a sense of each country’s level of commitment to try to limit global warming, I created a simple dataset with data from Climate Action Tracker, which rates countries based on Intended Nationally Determined Contributions (INDCs), 2020 pledges, long-term targets, and current policies, and whether they are consistent with a country’s fair share effort to the Paris Agreement 1.5°C temperature goal.
The rankings are as follows:
- 1.5°C Paris Agreement Compatible – Commitments are consistent with the Paris Agreement’s 1.5°C limit.
- 2°C Compatible – Commitments are consistent with the 2009 Copenhagen 2°C goal but not fully consistent with the Paris Agreement.
- Insufficient – Commitments are not consistent with holding warming below 2°C, much less with the Paris Agreement’s 1.5°C limit. If all targets were in this range, warming would reach over 2°C and up to 3°C.
- Highly Insufficient – Commitments are not at all consistent with holding warming to below 2°C. If all targets were in this range, warming would reach between 3°C and 4°C.
- Critically Insufficient – Commitments are not at all consistent with holding warming to below 2°C. If all targets were in this range, warming would exceed 4°C.
I wanted to make this chart very simple, especially after the rather complex map and, to a lesser degree, stacked graph. Again, the colors used for this chart are consistent with those used for the map and graph.
Dashboard design
In thinking about the design for the dashboard, I followed Few’s guidelines “to arrange the data in a manner that fits the way it’s used” and to make the “the most important data…prominent” (11). I wanted to organize the dashboard in a way that would allow users to start with the most complex map data related to the present situation, move to the slightly less complex graph data that provides a bit of a historical context, and finally end on the simplest data related to future commitments (although obviously, overall, users will interact with the visualizations in a much more dynamic, non-linear way). As such, the “Present” map visualization was placed on the left half of the dashboard (with its filter options and legend below) to function as an entry point, as our (Western) eyes initially gravitate to the top left when reading. The “Past” stacked graph was placed at the top right of the dashboard (with its filter options below, to be consistent with the map), as the next “natural” visual step for users. Lastly, the “Future” chart was placed at the bottom right of the dashboard as, in theory, the last (and “least important”) element users would engage with. To unify the dashboard, users are able to use each visualization as a filter for the others – clicking on a particular country will isolate the data for that country in all of the visualizations. This ability to filter and isolate data across visualizations also helps mitigate the potential difficulty with accurately interpreting individual trends in the stacked graph
The dashboard was designed to be viewed on a computer screen, and, at this time, is not mobile or tablet compatible; this is generally consistent with the primary target audience (professors might use this as a teaching tool in the classroom and students are more likely to do work/research on computers). Furthermore, it was designed so that the entire dashboard (at least the part with the visualizations) can be seen on one screen, with no need for scrolling. This is important because “not only does [it] provide convenience for viewers and save them valuable time, it also paints a complete picture that can bring to light important connections that might not be visible otherwise” (Few, 4). It is especially critical for these visualizations, which complement each other and have the ability to filter each other, and which collectively tell an overall narrative.
Despite the assumed background knowledge of my primary target audience, in order to make the visualization accessible to a wider audience as well, I also included further information and details below the dashboard in a text box. Furthermore, since all data and visualizations are not neutral or objective (as Sula points out “there may be residual biases in measurement design, modelling techniques or background assumptions…data collection is really data selection…” (31)), I wanted to be as transparent as possible about data sources/calculations to allow users to make their own judgements and draw their own conclusions.
User Experience Research Methods
The four participants I chose for user research more or less represented my primary and secondary target audiences (and therefore, hopefully, actual users), and I spoke with each of them at different stages of the design process. The first was a professor of environmental ethics (i.e. someone with significant background knowledge who might use such a visualization as a teaching tool), whom I initially interviewed before I started to work on the visualization in order to learn more about what would make this kind of tool most useful for his purposes. Another professor of environmental ethics was my second participant, with whom I also did a think-aloud and a follow-up discussion at an early stage in the design process (after the first iteration of the map visualization was complete). After I incorporated this feedback and once I had completed my initial overall design with all three visualizations, I got feedback from an older individual who is perhaps not as technologically adept but who has some general background knowledge and some interest, representing “the general public.” After a first round of revisions on the overall design, I did a think-aloud with a fourth individual whose lower background knowledge but high interest mirrored that of a potential student. I did a second round of revisions on the overall design based on this feedback, and I did a final think-aloud and follow-up discussion with the first participant.
Discussion/Findings
With this Tableau dashboard, I have created three independent but complementary visualizations that inform each other and collectively present the user with a comprehensive overall narrative while also allowing her/him/them to explore specific details more deeply. Users enter through the map, which offers a layered but succinct snapshot of the current situation – it is clear that there is a link between overall wealth and primary responsibility for emissions, which tend to be concentrated in the US, China, and Europe, while vulnerability tends to be concentrated broadly in Asia and Africa. Next, users are presented with a historical view of emissions, which shows that the current top emitters have generally consistently dominated for roughly the past sixty years (adding to their current responsibility), but that also shows that the total emissions per country somewhat obscures how it breaks down per capita (e.g. Qatar prominently pops to the top in the latter). With all of this information about the present and past, users can then see how these countries, especially those with high responsibility, fare in terms of their commitments to addressing the climate crisis.
While relatively complex, it does not seem to be overwhelmingly so based on user testing; just enough information is provided in the visualizations for the user to be able to extract meaning and draw conclusions, and redundancy is avoided, while additional (but not absolutely necessary) information is provided below if desired. However, many adjustments have had to be made along the way; below I discuss feedback from user research participants that helped refine my design and that offers possibilities for future directions.
Edit Data Included: Simplify and Add
During the early stages of the design, most feedback related to either simplifying the existing data to ensure that only what was relevant to the overall narrative – and nothing extraneous – was included, or adding to the existing data to add more dimensions and complexity to the dashboard. For example, on the first iteration of the map, the GHG emissions were broken down by gases (e.g. CO2, CH4, N20, etc.), but after the first user testing session, it became apparent that, while this information was interesting, it was not necessary for or directly relevant to the overall narrative. Furthermore, initially the map was the only visualization. Feedback from the first participant prompted me to include data on historical emissions as well as data on Paris Agreement commitments to complement the map visualization; this is what ultimately led to the overall past/present/future narrative. This also led to highlighting the top emitters – before this, all of the countries were one color. It was also suggested that I include GPD per capita and emissions per capita on the map, but, as mentioned above, the limits of the program (or my knowledge of how to use the program) essentially made this unfeasible, and I had to find a compromise by including emissions per capita in the historical graph.
Clarify with Legends, Labels, Annotations, and Headings
In the intermediate stages of designing the dashboard – after all three visualizations were included but it still needed to be refined – most feedback related to the need for clarifying elements. For example, before I settled on the current color palette, feedback from the “general public” participant indicated that it was not immediately clear what the colors meant or which countries they were associated with. This prompted me to include a legend and to change the color palette.
The labels for the map filters also went through several iterations based on participant feedback (primarily from one of the professors and the “student” participant). Initially, they did not include any heading (e.g. “Context,” “Responsibility,” and “Vulnerability”) but were simply descriptive. The heading for “Context” was also initially “Background” which was deemed “confusing.” Eventually, I settled on the current labels, which hopefully strike a balance between succinctness and description.
The “student” and “general public” participants also asked for more clarifying information for the units of measurement (e.g. MtCO2) as well as for the different kinds of emissions (territorial and consumption). Ideally I would have liked to include pop-up annotations that would appear when a user hovered over these elements, but I was unable to find a way to do this; as such, this issue was addressed primarily by including further information in the text below the dashboard.
Lastly, the headings for each visualization went through several iterations as well. In particular, the “student” participant found the initial heading for the future section (“Future: Commitment Rankings”) to be unclear. As such, each heading was adjusted to succinctly but clearly introduce the user to what kind of information they should expect to be looking at in each visualization.
Future Directions
Although I have already made revisions to the dashboard based on results from user testing, there are still several possible future directions and considerations for this project. At the data level, I would probably want to do an even more thorough compilation of data to ensure the numbers are as up-to-date and accurate as possible. I would also attempt to address the issue of lack of data for several countries in certain filters in a way that was more thorough, systematic, and consistent.
I would also consider expanding on the current visualizations or adding to them. It would be useful to be able to include emissions per capita on the map, as mentioned above, if it is somehow possible to assign separate scales to each filter on the map. There are also several other possible measures for responsibility and vulnerability that could be considered (e.g. the effects on agriculture or on non-human animals).
Finally, I would be interested in further developing the “Future” section to be a little more nuanced and comprehensive – only a handful of countries are included because of the source. Other suggestions from user testing that might be considered include incorporating a country profile to demonstrate how historical interventions have contributed to their current socio-economic and political statuses.
References
Few, S. (2009). Now you see it: Simple visualization techniques for quantitative analysis. Oakland, California: Analytics Press.
Heer, J., Bostock, M., & Oglevetsky, V. (2010). A tour through the visualization zoo: A survey of powerful visualization techniques, from the obvious to the obscure.” ACM Queue 8(5).
MacDonald, L. (1999). Using color effectively in computer graphics. Computer Graphics and Applications, IEEE 19(4), 20-35.
Sula, C.A. (2013). Quantifying culture: Four types of value in visualization. In J.P. Bowers, S. Keene, & K. Ng (Eds.), Electornic Visualisation in Arts and Culture (25-37). Springer.